This study aimed to determine the effectiveness of a preferred intensity exercise intervention on the depressive symptoms of adolescents receiving treatment for depression at post-intervention (week 6) and six months follow-up. No effect on depressive symptoms was found at post-intervention, however a statistically significant effect on depressive symptoms was found at six-month follow-up in favour of the intervention.
Previous trials investigating the impact of exercise on depressive symptoms in adolescents have reported mixed results, with some reporting statistically significant treatment effects for depressive symptoms [34–36] and other studies reporting no such treatment effects [37–39]. The current study adds to this body of literature, suggesting no effect of exercise as an additional treatment alongside TAU immediately post intervention.
The likely reason for the non-significant antidepressant effect is that the study is lacking the statistical power required to detect a difference. At the outset, using Cohen parameters , a medium effect size was anticipated. As a consequence, our power calculation suggested 128 participants were required in order to have 80 % to detect such a difference. Only 87 participants were recruited, and as such, the difference may have been missed.
A further possible explanation for the lack of a statistically significant treatment effect is that this trial was, in essence, attempting to determine the additional benefit of exercise alongside TAU. No previous studies have investigated the added benefit of exercise for a clinical sample of adolescents receiving mental health treatment. As such, the small effect observed in previous trials may be due to the comparison treatment being substantially less efficacious.
Another potential contributor to the non-significant treatment effect refers to the six-week duration of our intervention; this duration may be viewed as short or medium-term compared to: i) the advice from the National Institute for Care Excellence (NICE) (2005) of a ten-twelve week long structured exercise programme for children and adolescents with depression and ii) a recent RCT of a 12 week exercise intervention reporting antidepressant treatment effects . Moreover, trial participants in this study were required to exercise at their preferred intensity. Given the low self-esteem and self-efficacy levels of people with depression, a number of sessions were used by trial participants to explore and eventually select preferred intensity before building up successful experiences and positive affective responses. Thus, an intervention with a longer duration may have led to higher treatment effects.
Our concomitant qualitative study  in which we interviewed 26 participants who completed the intervention arm of the trial revealed that a number of participants’ mood dropped following completion of the intervention. They stated this was due to losing something that improved their mood, provided a distraction, increased their self-efficacy, and improved overall motivation. In contrast, the control arm participants continued TAU and therefore were unlikely to have experienced the loss of a structured intervention, at least not at the time of the post intervention measures. It is proposed that the disappointment and dip in mood experienced as a result of ending their participation in intervention may have temporarily masked any improvement in depressive symptoms.
Considering the statistically significant decrease in depression scores from baseline for those in the intervention group alongside the confidence intervals of the between group difference suggesting that the real difference may be up to seven points in favour of the intervention. The findings suggest that exercise may still hold promise as a treatment for depression in this population.
Depressive symptoms at six months
A significant treatment effect was observed at six month follow up whereby allocation to the treatment arm was predictive of approximately a five point difference in depressive symptoms, compared to the control arm. The intervention appeared to have a delayed effect on depressive symptoms.
This treatment effect at six months is a particularly novel finding as the majority of previous studies investigating the impact of exercise on depression in adolescents have not included long term follow ups. There have, however, been two notable exceptions. In a large RCT comparing exercise alongside a 50 min educational and cognitive behavioural therapy (CBT) class compared to an equivalent contact, no exercise comparison for high school students, Melnyk et al.  found no statistically significant differences between groups on depression scores at six months. However, this may be explained by a potential floor effect owing to the low depressive symptoms observed at baseline.
Conversely, Hughes et al.  investigated the impact of a 12 week exercise intervention on the depressive symptoms of adolescents diagnosed with MDD, and included follow ups at 26 and 52 weeks. No statistically significant differences were found on depressive symptoms at either time point. However, the small numbers analysed at 26 weeks (10 vs 9) and at 52 weeks (7 vs 8) potentially explain this lack of treatment effect. In contrast, the statistically significant effect observed in our trial at six months, possibly stems from the larger sample.
There were no significant demographic or clinical differences between participants who were lost to follow up and participants who remained in the study, consequently, it appears that the improvement in depressive symptoms observed at six months may have been attributable to engagement with the exercise intervention. However, no between group differences were observed on the LTEQ at six months, indicating that those in the intervention arm did not continue to exercise above and beyond control participants. As such, it is likely that the mechanism of action may be the additional positive experiences of the intervention reported in the concomitant qualitative study  as opposed to increased exercise.
The improvements in depressive symptoms occurring six months post-intervention concurs with the premises of the Transtheoretical Model of Change by Callaghan et al. . In this study it is reported that significant behaviour changes take at least six months to take effect. The time-demanding aspect of the exercise intervention (exploration and selection of the preferred intensity exercise before building up successful experiences and positive affective responses) supports the conclusion Callaghan et al. reported.
When viewed in context, the treatment effect is interpreted as an additional decrease of five points on the CDI-2 over TAU only. The difference in modelled scores from baseline to six-month follow up for the TAU only group was 3.8 points. This suggests that the added effect of an exercise intervention alongside TAU is capable of substantially increasing the depressive reducing effect of TAU alone. Furthermore, when viewing the confidence intervals of the between group comparison, the intervention may lead to a nine point reduction in CDI-2 points. This is likely to be considered clinically meaningful when considering the cut off scale for clinical symptoms of depression is 14 .
Health related quality of life (HRQOL)
No differences were observed between arms on HRQOL as measured by the EQ5D-5 L at either time point. To the authors’ best knowledge, no previous trials in this research area included measures of HRQOL. Consequently there is little data with which to compare the current findings. Nevertheless, there have been studies investigating the impact of exercise on quality of life (QOL) in adults. Importantly, it is acknowledged that HRQOL is not as broad a concept as QOL as its focus is on an individual’s health or disease status opposed to non-health related features of life as well. In a recent systematic review  on the impact of exercise on QOL for depressed adults and older adults, exercise was found to improve some QOL domains; primarily the Physical and Psychological domains.
There is some evidence that QOL in depressed adolescents can be improved by current treatment options. For instance, Vitiello et al. , found through a large RCT comparing CBT, fluoxetine and a placebo on adolescent depression, a positive effect on QOL for combined CBT and fluoxetine. In light of these findings, we anticipated that a possible explanation for the non-significant difference in HRQOL may be the relatively small sample.
An average attendance of 70 % and drop out of 8 % for participants who attended at least one session, suggest that the intervention was highly acceptable, especially since it was implemented in real life settings (community centres) and no external motives (e.g. vouchers) were provided to participants for attendance. Similarly high adherence rates have been reported by Hughes et al.  who conducted an RCT investigating an exercise intervention with depressed adolescents. In this trial, however, participants were given a $25 incentive per session attended, and despite this benefit, adherence typically reduced over time. The high acceptability of our intervention is further supported by the associated qualitative study  where the preferred intensity aspect of the intervention is highlighted as one of the key contributing factors regarding intervention adherence.
The importance of preferred activity is further evidenced through analysis of heart rate and perceived exertion data. A mean percentage of maximum heart rate of approximately 50 % and a mean RPE value of approximately 10 suggest that the participants preferred to exercise, on average, at a low intensity. Importantly, the norms through which the RPE is validated were developed for healthy populations . As such, an RPE value of 10 is likely to have a different meaning for this group than the general population. This is considered in light of depression typically being associated with a series of physical symptoms such as aches, tiredness, back pain and gastrointestinal problems . Moreover, the participants in this study were faced with additional physical symptoms due to self-harm and/or other medical conditions (e.g., sciatica or knee joint injuries). Thus, the selected low intensity exercise was unsurprising and through encouraging preferred intensity exercise, the participants did not experience injuries or adverse effects and tended to continue attending the intervention despite the various physical comorbidities seen in the sample.
This study is one of few well-designed trials that have tested the effect of preferred intensity exercise on depressive symptoms in a clinical population of adolescents. The study is a pragmatic RCT, therefore minimum exclusion criteria were employed and the intervention was delivered in a setting reflecting clinical practice. As such, this study has high external validity, as the included participants represented a clinical population, and the intervention was delivered in a ‘real life’ setting. The pragmatic design also allowed for a TAU control condition, thereby reflecting current clinical practice. Consequently, this study is the first to determine the added benefit of an exercise intervention to mental health TAU for depressed adolescents. Our findings, therefore, are of particular interest for researchers and practitioners involved in primary care. Moreover, our study recruited more participants than any previous trial of adolescents using mental health services for the treatment of depression. A six month follow-up period is also rare in studies of this nature with depressed adolescents. The use of preferred intensity exercise applied to populations seeking mental health care and treatment has been pioneered by our group  and this is the first such study targeting adolescents living with depression and seeking treatment from mental health services.
Despite being the largest study testing the effect of exercise in depressed adolescents, we were unable to recruit to the required sample size, and this may explain the lack of statistical significance at post-intervention. Moreover, the relatively small sample size increases the risk of a type 1 error which should be considered when interpreting the six month treatment effect. The failure to recruit to the required sample size was due to difficulties in engaging CAMHS and G.P services at the outset of the project, however once the study had been promoted within all the relevant teams, recruitment increased substantially. However, funding was not available to continue data collection beyond the pre-specified end point.
Further limitations may include the relatively short duration of the intervention, the inclusion of exercising young people and the lack of data concerning the amount of exercise being undertaken by participants during the intervention period.